On Mon, 10 May 2004, Stephen Burke wrote:
> On Mon, 10 May 2004, Ian Stokes-Rees wrote:
> > If so, is it correct to assume that we (LHCb) should try to keep our
> > jobs to under 12h CPUtime and 24 hours wall clock time so they can fit
> > in the default "long" queues, rather than be relegated to the (possibly)
> > lower priority/smaller resource "infinite" queues?
>
> The experiments can negotiate to have the queues set up in a way which
> suits them, I don't think there's anything magic about 12 hours. For
> sites which mostly support a single VO they should be prepared to do it
> however you want. Other sites will have to do a bit of arbitration, but
> probably the needs of the experiments aren't so different that you can't
> get a reasonable compromise. Potentially you can have VO-specific queues
> even if the resources they feed are shared.
True to a point but sites won't be that willing to have extremely long
queues. I clearly have to drain nodes to install a kernel.
Generally just put in your JDL what you feal you need and don't look
at the pbs level of things. If you find you are hardly getting any resources
in a job list match then ask sites why they won't support you.
Steve
>
> Stephen
>
--
Steve Traylen
[log in to unmask]
http://www.gridpp.ac.uk/
|